Skip to content

Add node selector configuration for Python SDK (katib.tune)Β #2603

@hahahannes

Description

@hahahannes

What you would like to be added?

It would be nice to be able to specify on which node Katib jobs would run. This corresponds basically to the node_selector field of a Pod.

For example:

katib_client.tune(
    name=name,
    objective=objective,
    parameters=parameters,
    objective_metric_name="loss",
    max_trial_count=5,
    resources_per_trial={"cpu": "2"},
    node_selector="nvidia.com/gpu.product": "NVIDIA-H100-NVL"}
)

https://github.com/kubeflow/katib/blob/master/sdk/python/v1beta1/kubeflow/katib/api/katib_client.py#L179

Why is this needed?

If you have a heterogeneous cluster with node configs, e.g. different GPU vendors, models, ... a user might want to specify which node should be used.

Love this feature?

Give it a πŸ‘ We prioritize the features with most πŸ‘

Metadata

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions